Low Level Virtual Machine

Low Level Virtual Machine
source
Name	Low Level Virtual Machine
Developer	LLVM Project
Released	2003
Latest release	LLVM 17.0.6
Programming language	C++
Operating system	Cross-platform
License	University of Illinois/NCSA Open Source License

Low Level Virtual Machine Low Level Virtual Machine is a modular compiler infrastructure project used to build compilers, toolchains, and runtime systems. It provides a reusable set of libraries and tools, an intermediate representation, and optimization passes that enable language implementers, operating system developers, and hardware designers to produce performant native code. The project is developed and maintained by a broad community including contributors from universities, corporations, and standards organizations.

Introduction

The project originated as a research effort at University of Illinois at Urbana–Champaign and evolved into a large open-source ecosystem involving organizations such as Apple Inc., Google LLC, Microsoft, Intel Corporation, and ARM Ltd.. Its modular design influenced work at institutions like Stanford University, Massachusetts Institute of Technology, and Carnegie Mellon University and is referenced in standards discussions at bodies including ISO and IEEE. Popular products and languages such as Clang (compiler), Rust (programming language), Swift (programming language), Julia (programming language), and Kotlin rely on its components or interoperate with its intermediate representation. The project's governance includes contributors from research labs, commercial vendors, and independent developers collaborating via repositories hosted by organizations like GitHub and processes influenced by communities such as The Linux Foundation.

History and Design Goals

The original design was led by researchers including Chris Lattner while at University of Illinois at Urbana–Champaign and later influenced by work at Apple Inc. and other industry labs. Initial goals emphasized a flexible, language-neutral infrastructure that could support aggressive optimizations and retargeting to different instruction set architectures such as x86, ARM, PowerPC, and RISC-V. The roadmap and governance adapted through contributions from corporate engineering groups at Google LLC, NVIDIA Corporation, AMD, and Intel Corporation and academic collaborations with University of Cambridge and ETH Zurich. The project’s licensing and stewardship were affected by interactions with institutions like Open Source Initiative and legal teams from major contributors.

Architecture and Components

The architecture centers on a typed, SSA-based intermediate representation and a set of modular libraries for analysis, transformation, and code generation. Core components include frontend tools such as Clang (compiler), middle-end optimizers used by teams at Mozilla and Red Hat, and backends for targets including x86-64, ARM64, PowerPC64LE, and RISC-V. Ancillary projects and libraries such as LLDB, LTO, polly, and libc++ integrate with runtime systems like Android (operating system), iOS, and Linux kernel development environments. Tooling for debugging and profiling connects to ecosystems maintained by Valgrind, GDB, and vendor suites from Intel Corporation and NVIDIA Corporation.

Intermediate Representation and Language Features

The intermediate representation was designed to be language-agnostic and suitable for both high-level optimizations and low-level code generation; it supports features pioneered in compiler research at Princeton University, University of California, Berkeley, and University of Illinois at Urbana–Champaign. It adopts static single assignment form influenced by work from IBM Research and contains constructs for memory, control flow, and metadata used by projects such as Clang (compiler), Rust (programming language), and Swift (programming language). The IR and associated bitcode format are consumed by tools in the ecosystem including linkers used by GNU Project toolchains and runtime systems like LLVM's JIT employed in runtimes such as V8 (JavaScript engine) and Mono (software). Language features such as intrinsic functions, attributes, and metadata support interoperation with standards bodies like POSIX and hardware descriptions from ARM Ltd..

Toolchain and Ecosystem

The toolchain comprises frontends, optimizers, backends, linkers, and runtime support libraries. Prominent frontends include Clang (compiler), Rust compiler (rustc), Swift compiler (swiftc), and experimental frontends developed at Google LLC and academic groups. Build systems and continuous integration frameworks from organizations such as CMake, Bazel, and GitLab host integration points, while package managers and distribution projects like Debian, Fedora, Homebrew (package manager), and npm facilitate deployment. Commercial vendors including Apple Inc., Google LLC, Microsoft, NVIDIA Corporation, and AMD provide tested binaries and platform-specific enhancements. Research tooling and education use cases arise from collaborations with Coursera, edX, and university courses at Stanford University and MIT.

Applications and Use Cases

Adopters range from systems and application developers to academic researchers. Major operating systems and platforms—macOS, iOS, Android (operating system), Windows, Linux (kernel), and embedded RTOS projects—use compilers or components built on the project. Language implementers for Rust (programming language), Swift (programming language), Kotlin, Julia (programming language), and Haskell toolchains utilize its optimizers and code generators. High-performance computing centers and cloud vendors such as Amazon Web Services, Google Cloud Platform, and Microsoft Azure leverage backend code generation and link-time optimization for workloads in projects like TensorFlow, PyTorch, LLVM-based JITs, and vendor libraries from NVIDIA Corporation and Intel Corporation.

Performance, Security, and Criticisms

Performance engineering efforts draw on benchmarks and techniques from collaborations with SPEC, Phoronix, and academic research at ETH Zurich and University of Cambridge. Security analyses and hardening efforts have involved teams from Mozilla, Google LLC, and Microsoft focusing on mitigations for speculative execution issues publicized alongside researchers at University of California, Berkeley. Criticisms include debates over governance, complexity, and binary size raised by contributors and distributions such as Debian and Fedora as well as discussions in communities like LLVM-dev and forums hosted on GitHub. The project continues to evolve through community proposals, corporate sponsorship, and academic research to address scalability, verification, and portability concerns.

Category:Compilers Category:Free and open-source software projects Category:Software using the NCSA/University of Illinois license